R/rowdiff.R

In manymodelr: Build and Tune Several Models

#' Get row differences between values
#' @description This function returns the differences between rows depending on the user's choice.
#' @param df The data set for which differences are required
#' @param direction One of forward and reverse. The default is forward meaning the differences
#' are calculated
#' in such a way that the difference
#' between the current value and the next is returned
#' @param exclude A character vector specifying what classes should be removed. See examples below
#' @param na.rm Logical. Should missing values be removed? The missing values referred to are those introduced during the calculation ie when subtracting a row with itself.
#' Defaults to FALSE.
#' @param na_action If na.rm is TRUE, how should missing values be replaced? Depending on the 
#' value as set out in `na_replace`, the value can be replaced as per the user's requirement.
#' @param ... Other arguments to `na_replace`.
#' @return A data.frame object of row differences
#' @examples
#' # Remove factor columns
#' data("yields", package="manymodelr")
#' rowdiff(yields,exclude = "factor",direction = "reverse")
#' rowdiff(yields[1:5,], exclude="factor", na.rm = TRUE, 
#' na_action = "get_mode",direction = "reverse")
#' @seealso \code{\link{na_replace}}
#' @export
rowdiff<-function (df, direction = "forward",exclude=NULL,na.rm=FALSE,na_action=NULL,...){
  UseMethod("rowdiff")
}

#' @export
rowdiff.data.frame<-function(df, direction = "forward", exclude=NULL, na.rm=FALSE,na_action=NULL,...){
  
  if(!is.null(exclude)){
   
    df<-Filter(function(x) ! class(x) %in% exclude, df)    
  }

  stopifnot("Only forward and reverse are supported"= direction %in% c("forward","reverse"))
  
 if(direction=="forward"){
    
   res<-as.data.frame(sapply(df,function(x) x-dplyr::lead(x,1)))
    
}

  if(direction=="reverse"){
    res<-as.data.frame(sapply(df, function(x) x-dplyr::lag(x,1)))
  }

  if(na.rm){
      res<-na_replace(res,how=na_action,...)
      
}
   
res



}